SPATIAL STATISTICS ON AMAZON RAINFOREST ASSESSMENT: SPATIALLY STRATIFIED INVENTORY PROCESSING

Abstract

Biomass and wood volume estimates in forest ecosystems are fundamental to a variety of studies focusing at forest dynamics. These estimates are usually carried out through forest inventory techniques which rely upon statistical computations. This work aims at providing a new methodological approach to forest inventory processing when data is georeferenced. Specifically, geostatistical modelling is performed through ordinary co-kriging using tree basal area and tree richness as a cofactor in an Amazonian rainforest site. The spatial interpolation provided the tools for the creation of two disjoint forest strata, which are processed following the principles of Stratified Forest Inventory. The spatially stratified forest inventory processing has shown a 14.29% decrease in error as directly compared to simple random sampling processing. Only two strata have been used following spatial interpolation, albeit it is argued that theoretically any number of them could be generated. The procedure is methodologically feasible and offers a framework to future research on its development and reach. Particularly, the geometries of forest strata and the behavior of spatial interpolation along a gradient of forest vertical structures are of potential interest in future work.